SOT-227(ISOTOP) vs TO-264 heat dissipation

[Fauks]

I'm working on making an electronic load and have narrowed it down to using a linear MOSFET in a SOT-227 module or a TO-264 package.

I would assume the module has better heat dissipation with better connections for thick wires vs using PCB traces. But according to the two data sheets (IXTK200N10L2, IXTN200N10L2) the TO-265 is rated a bit higher and also a bit cheaper. 

SOT-227 module:
SOA = VDS:100V, ID:5A, TC:75°C , tp:5s, 500W

TO-264:
SOA = VDS:100V, ID:6.25A, TC:75°C, tp:5s, 625W

Anyone with experience with these two packages? I'm still leaning toward the SOT-227 but I thought I would ask first.

Thank you!

[T3sl4co1l]

Definitely the SOT-227: it's isolated, so can be put down with grease alone, and the heatsink grounded.  They have a DBC (direct bonded copper) isolator in there; some manner of ceramic, probably AlN, with copper deposited directly onto its surface as the name suggests.  This is present by default in module packages, I believe; it's rarely present in transistors otherwise, though does show up fairly often among rectifiers and thyristors.

But why not five TO-247s instead?  Of SJ types of suitable SOA, or QFETs if you can still get some.  Which way is cheaper?  Do you need the low saturation voltage, could you ballast up some or much of the range with resistors (in series, switched in parallel, etc.)?  There are many ways to slice this.  Resistors have the additional effect of reducing short-circuit current available to the transistor, in case the control goes batty, or the transistor fails catastrophically.

Tim

[coppercone2]

the module has so many advantages like being easy to replace without messing with a heavy board or whatever you figure out.

if you can use wire then use it

heatsink on board sucks anyway if you can get away from it get away from it. You get so much flexibility in mechanical design from having a wire

They like TO-264 in premium power electronic solutions (i.e. non budget welders).

[tooki]

I'm working on making an electronic load and have narrowed it down to using a linear MOSFET in a SOT-227 module or a TO-264 package.

I would assume the module has better heat dissipation with better connections for thick wires vs using PCB traces. But according to the two data sheets (IXTK200N10L2, IXTN200N10L2) the TO-265 is rated a bit higher and also a bit cheaper. 

FYI, you posted two different screenshots of the same datasheet, not of the two different ones.

[wraper]

heatsink on board sucks anyway if you can get away from it get away from it. You get so much flexibility in mechanical design from having a wire

People put heatsinks along the edge of the board all the time just fine. In fact it's way less job mounting like that if done properly rather than running 3 or 4 wires with lugs crimped. IMHO using SOT-227 makes sense in something like industrial equipment cabinet with modules wired together and it being used like a module. It makes very little sense inside a device that is not supposed to be serviced in the field.

[coppercone2]

its easier to put together once you make 10 but thats about it. the amount of design flexibility you get from not having heat sinks in specific orientations on PCB is huge. The smaller your boards are and the more wiring you have the easier it is to move something.

Miller uses them in their ultra compact welders, its good.


You can drill out a large lug with 4 screws to mount directly to the 264 package also. You can even cut them out like a pizza to make contact with 3 terminals,or two terminals using only one wire with no reliability problems, so long the ring lug is heavy gauge.

[wraper]

You can drill out a large lug with 4 screws to mount directly to the 264 package also. You can even cut them out like a pizza to make contact with 3 terminals,or two terminals using only one wire with no reliability problems, so long the ring lug is heavy gauge.

That would not be a reliable connection though. You don't want FR-4 to be directly screwed through. There are several issues with that, one of them is that FR-4 may compress and loosen over time. Not to say such packages often do not guarantee that terminals are exactly on the same level with each other. BTW you can bend TO-264 legs by 90^o to the front and mount it on heatsink under PCB, add a hole in PCB for screw access.

[coppercone2]

a lug is the copper crimp you put on a bolt

[coppercone2]

however miller also did put PCB on them and they did also have the bright idea to clamp down on solder covered braids and they don't know everything.

A big via might help for a bolt down PCB but that is electrolytically formed copper that is kinda brittle. I imagine you would want to put a heavy bronze stud on it that is soldered on both sides (taller then PCB) which the bolt goes through and tries to compress.

[Fauks]

I'm working on making an electronic load and have narrowed it down to using a linear MOSFET in a SOT-227 module or a TO-264 package.

I would assume the module has better heat dissipation with better connections for thick wires vs using PCB traces. But according to the two data sheets (IXTK200N10L2, IXTN200N10L2) the TO-265 is rated a bit higher and also a bit cheaper. 

FYI, you posted two different screenshots of the same datasheet, not of the two different ones.

Oops! I have uploaded the correct datasheet screenshot.

You can drill out a large lug with 4 screws to mount directly to the 264 package also. You can even cut them out like a pizza to make contact with 3 terminals,or two terminals using only one wire with no reliability problems, so long the ring lug is heavy gauge.

That would not be a reliable connection though. You don't want FR-4 to be directly screwed through. There are several issues with that, one of them is that FR-4 may compress and loosen over time. Not to say such packages often do not guarantee that terminals are exactly on the same level with each other. BTW you can bend TO-264 legs by 90^o to the front and mount it on heatsink under PCB, add a hole in PCB for screw access.

I was planning on making a dedicated aluminum based PCB for the module/package (upside down) and mounting a large CPU heatsink on top of it, using the mounting bracket supplied with the heatsink (standard AM4/5 or xTR4 for two fets)  The PCB would be solely for mounting purposes. I would have a mounting bracket cut if it was more affordable compared to an aluminum based PCB.

I could also design and print something maybe.

Definitely the SOT-227: it's isolated, so can be put down with grease alone, and the heatsink grounded.  They have a DBC (direct bonded copper) isolator in there; some manner of ceramic, probably AlN, with copper deposited directly onto its surface as the name suggests.  This is present by default in module packages, I believe; it's rarely present in transistors otherwise, though does show up fairly often among rectifiers and thyristors.

But why not five TO-247s instead?  Of SJ types of suitable SOA, or QFETs if you can still get some.  Which way is cheaper?  Do you need the low saturation voltage, could you ballast up some or much of the range with resistors (in series, switched in parallel, etc.)?  There are many ways to slice this.  Resistors have the additional effect of reducing short-circuit current available to the transistor, in case the control goes batty, or the transistor fails catastrophically.

Tim

I'm really new to all of this. This is my first bigger project. From what I could gather, multiple smaller fets in parallel was going to be more difficult than running larger packages with higher ratings. Unfortunetly, there are not many linear fets to choose from.

I'm looking to scale up the design from the "MightyWatt R3", using the same control circuit but with a larger load capability. I have attached the schematic I am referring to.

[tooki]

a lug is the copper crimp you put on a bolt

Crimped or soldered, both are common (though solder lugs are definitely far less common these days).

[Fauks]

Here is a little mockup render to show what I am trying to accomplish.

The mounting holes are standard AMD AM5 socket heatsink mounting. The thermal pad under the MOSFETs is 3mm thick which brings the heatsink plate to the same height that a AM5 mounted CPU cooler would sit. So the top case side would get some cooling through the thermal pad and the top heat sink plate would get cooling from the tower cooler. It's pretty easy to find coolers that can handle ~280w continuous and for short bursts much higher. 

Solder pads for source/drain would be a few mm away from the terminals so keep trace length as short as possible. The PCB would be slightly larger to accommodate mounting holes for some sort of case. Maybe a mini ITX computer case? Also would house a separate small control board. 

I've also attached a mock up of how I was thinking of using the modules, and also how the heatsink would mount.

Thoughts?

[T3sl4co1l]

I think you'll be disappointed using thru-hole parts with an aluminum PCB... check the specs from the fab carefully.  Isolated thru-holes will cost extra, or you need to find someone that does it.

I'd be fine with a brace cut from hardware store metal stock.

Tim

[Fauks]

I think you'll be disappointed using thru-hole parts with an aluminum PCB... check the specs from the fab carefully.  Isolated thru-holes will cost extra, or you need to find someone that does it.

I'd be fine with a brace cut from hardware store metal stock.

Tim

I didn't even think about that. I guess FR4 would be fine. The heatsink I have in mind has a pretty solid steel back plate to brace the board for the sink. Maybe use 2oz copper as well.

[tszaboo]

I'm working on making an electronic load and have narrowed it down to using a linear MOSFET in a SOT-227 module or a TO-264 package.

I would assume the module has better heat dissipation with better connections for thick wires vs using PCB traces. But according to the two data sheets (IXTK200N10L2, IXTN200N10L2) the TO-265 is rated a bit higher and also a bit cheaper. 

SOT-227 module:
SOA = VDS:100V, ID:5A, TC:75°C , tp:5s, 500W

TO-264:
SOA = VDS:100V, ID:6.25A, TC:75°C, tp:5s, 625W

Anyone with experience with these two packages? I'm still leaning toward the SOT-227 but I thought I would ask first.

Thank you!

You add together the Rthjc and Rthcs values, whichever is lower will handle more power. How much? Not 500W, that only applies to the magical world of room temperature infinite heatsinks and junctions at their rated max tempertaure. Not surprisingly the SOT227 is better. SOT227 will be better than anything other than those exotic MOSFET modules with gigantic packages. Even if the datasheet will tell you otherwise, it's just a much more solid way of mounting the part on a heatsink, with more clamping force and better heat transfer.

[temperance]

The open loop output impedance of the OPA194 is 375 R. Driving a 23 nF load which increases towards 30 nF in the linear region with 375 R results in output stage bandwidth about 100 times lower in frequency the op amp. Or you have to add a high current buffer or properly compensate the op amp.

You also need an RC snubber circuit across the MOSFET or any wiring inductance will turn this circuit into a very good oscillator. 2.2...4.7 R with 100...470 nF will do up to some length of wire. (The oscillator is formed by the inductance connected to the drain and the MOSFET Cgd)

Wrong or faulty dummy load schematics are being posted on EEVblog almost every week.

A good post on the loop compensation subject made long ago by Jay_Diddy_B
https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/

Here I've done the same with a low cost op amp and some other low cost components:
https://www.eevblog.com/forum/beginners/constant-current-load-stability/25/
The schematic of a buffered op amp is in post 31
A test result is posted under reply 42

The remainder of what is being said is mostly people stating how I'm wrong and can be ignored. Except for what TopQuark has posted. (TopQuark, it somehow always makes me think of Dutch cheese? But the Top version.)

[coppercone2]

The board on transistor concept you have is liable to get loose. I don't know how to solve that problem. I recommend copper cable to lead it to a PCB where its terminated to a  ring lug mount (its like a bracket soldered to the PCB you screw into with a short jumper).

If you use through hole ring lug mounts, with a hole on the bottom, it MIGHT work if you had trimmed screws and a funky assembly procedure of inserting a washer sideways underneath the ring terminal mount. They would need to be aligned very well.

Now IIRC the miller welder that did this direct2pcb, it had OK screw tension (after 15 years), a bit low, on the PCB they had. But this will depend on your PCB quality (how good the epoxy is, and probobly other factors). I don't like it, they should have used a small wire adapter. That welder did explode on someone for no reason, possibly because one of their dodgy junctions gave up.

If you MUST have the pcb there, its probobly better to use the leaded package, with a hole in the PCB for mounting the part, or a bracket that compresses the part from both sides (better) using two screws



So unless you plan on using cable, go with picture #1, and put holes on the PCB so you can screw down the transistor module. This is not a bad design IMO. The only problem is the transistor might twist and yank on the leads. Optimally you would put notches and spacing between the transistors so you can grip them with a jig or pliers (i.e. duck bill pliers) when tightening so the part does not twist. Its a problem because you only got 1 mounting hole. In the instruction manual I would expect there be a page on how to replace the transistor properly (show pliers being used to restrain it)






***
I am curious about how a large eyelet could work for picture #3 to prevent the board from experiencing tension and loosening. This would be special hardware that is staked and soldered to PCB and it would need to be beefy, probobly made from bronze, assembled on a arbor type press and soldered using resistance techniques (lumo)

[Fauks]

The open loop output impedance of the OPA194 is 375 R. Driving a 23 nF load which increases towards 30 nF in the linear region with 375 R results in output stage bandwidth about 100 times lower in frequency the op amp. Or you have to add a high current buffer or properly compensate the op amp.

You also need an RC snubber circuit across the MOSFET or any wiring inductance will turn this circuit into a very good oscillator. 2.2...4.7 R with 100...470 nF will do up to some length of wire. (The oscillator is formed by the inductance connected to the drain and the MOSFET Cgd)

Wrong or faulty dummy load schematics are being posted on EEVblog almost every week.

A good post on the loop compensation subject made long ago by Jay_Diddy_B
https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/

Here I've done the same with a low cost op amp and some other low cost components:
https://www.eevblog.com/forum/beginners/constant-current-load-stability/25/
The schematic of a buffered op amp is in post 31
A test result is posted under reply 42

The remainder of what is being said is mostly people stating how I'm wrong and can be ignored. Except for what TopQuark has posted. (TopQuark, it somehow always makes me think of Dutch cheese? But the Top version.)

I appreciate the info. I found this schematic here:  https://github.com/kaktus85/MightyWattR3
It seems like a pretty popular open source electronic load I found when searching but it is now a few years old. I was planning on making changes and incorporating a small ESP32 instead of an old Arduino, among other changes.

It sounds like I have a lot more reading to do! 

The board on transistor concept you have is liable to get loose. I don't know how to solve that problem. I recommend copper cable to lead it to a PCB where its terminated to a  ring lug mount (its like a bracket soldered to the PCB you screw into with a short jumper).

If you use through hole ring lug mounts, with a hole on the bottom, it MIGHT work if you had trimmed screws and a funky assembly procedure of inserting a washer sideways underneath the ring terminal mount. They would need to be aligned very well.

Now IIRC the miller welder that did this direct2pcb, it had OK screw tension (after 15 years), a bit low, on the PCB they had. But this will depend on your PCB quality (how good the epoxy is, and probobly other factors). I don't like it, they should have used a small wire adapter. That welder did explode on someone for no reason, possibly because one of their dodgy junctions gave up.

If you MUST have the pcb there, its probobly better to use the leaded package, with a hole in the PCB for mounting the part, or a bracket that compresses the part from both sides (better) using two screws



So unless you plan on using cable, go with picture #1, and put holes on the PCB so you can screw down the transistor module. This is not a bad design IMO. The only problem is the transistor might twist and yank on the leads. Optimally you would put notches and spacing between the transistors so you can grip them with a jig or pliers (i.e. duck bill pliers) when tightening so the part does not twist. Its a problem because you only got 1 mounting hole. In the instruction manual I would expect there be a page on how to replace the transistor properly (show pliers being used to restrain it)






***
I am curious about how a large eyelet could work for picture #3 to prevent the board from experiencing tension and loosening. This would be special hardware that is staked and soldered to PCB and it would need to be beefy, probobly made from bronze, assembled on a arbor type press and soldered using resistance techniques (lumo)

I appreciate the reply.

The mounting pattern is a standard CPU socket mounting pattern (75mmx75mm, my quick render may not be to scale but it is close) and most heat sinks include mounting hardware including a steel brace that goes on the under side of the board, and two brackets that screw into that brace from the top to which the heat sink attaches.  I have attached an image of the hardware included in the heatsink I had in mind. It clamps down with quite a bit of force, so assuming the thermal pad is thick enough I don't see why it would come loose. It has springs that keep constant tension over time.

Using the SOT-227 module, I would have to use threadripper/Epyc compatible coolers as the heat sink base plate is much larger on those but the same principle would still apply. The current spacing is for LGA1700 sockets which is just standard intel 12th+ gen CPUs. 

Here is a short video showing how the mounting would work. I apologize if I am completely missing your point though lol. 
https://youtu.be/Ioaon-11z24?t=50

[T3sl4co1l]

Note that SOT-227 top side is not made to bear force; nuts are captive but loose in the package, and the leads are bent over tabs.  A mounting plate is best.  You might take a slab of hardware store extrusion and drill and tap it for the necessary features, or order a plate machined to the same.  I'm not sure offhand how that would work with the stock brackets; you may be better off finding some spots to drill and tap into the heatsink itself.

Tim

[Fauks]

Note that SOT-227 top side is not made to bear force; nuts are captive but loose in the package, and the leads are bent over tabs.  A mounting plate is best.  You might take a slab of hardware store extrusion and drill and tap it for the necessary features, or order a plate machined to the same.  I'm not sure offhand how that would work with the stock brackets; you may be better off finding some spots to drill and tap into the heatsink itself.

Tim

Interesting point. What about some aluminum spacers between the tabs and nut for bracing? Or like someone else mentioned, bending the tabs straight and mounting through the PCB so the nuts and making contact instead of the tabs.

Or are you referring to the top of the assembly in general should not bare weight?

[tszaboo]

Note that SOT-227 top side is not made to bear force; nuts are captive but loose in the package, and the leads are bent over tabs.  A mounting plate is best.  You might take a slab of hardware store extrusion and drill and tap it for the necessary features, or order a plate machined to the same.  I'm not sure offhand how that would work with the stock brackets; you may be better off finding some spots to drill and tap into the heatsink itself.

Tim

Interesting point. What about some aluminum spacers between the tabs and nut for bracing? Or like someone else mentioned, bending the tabs straight and mounting through the PCB so the nuts and making contact instead of the tabs.

Or are you referring to the top of the assembly in general should not bare weight?

SOT227 has two mounting holes on the base plate of it. It needs to be bolted down with these.

[coppercone2]

but for the other package you can either use a screw hole or put a bar across it that braces it from both sides. i mean that you need a way to screw it down aka a hole in the thing. and when you screw it in the part will twist so you need a way to hold it. your supposed to screw it in before soldering too but it would be good to be able to hold it nicely aligned, which means you need cutouts for pliers to go in or some kinda jig to hold it like a spanner. At least with grease it will, if you use a silpad maybe it stay put while tightening.


the other package has a thick base so screwing it down by the flange is the proper method. But you also wanna put holes there if you can, because otherwise you need specialty low profile screw drivers etc (annoying as fuck)

[coppercone2]

for drilling and tapping a heatsink you probobly want to consider using threaded inserts (helicoil) because aluminum and copper threads get damaged very easily and aluminum also makes bolts stuck easily

Usually heatsinks want a proper nut made out of steel